You are on page 1of 12

Engineering Structures 123 (2016) 213–224

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

An investigation on stressing and breakage response of a prestressing


strand using an efficient finite element model
A.B.M. Abdullah ⇑, Jennifer A. Rice, H.R. Hamilton, Gary R. Consolazio
Department of Civil and Coastal Engineering, University of Florida, Gainesville, FL 32611, USA

a r t i c l e i n f o a b s t r a c t

Article history: The mechanical response of a wire strand is inherently complex because the helical wires undergo
Received 11 March 2015 evolving stress and contact conditions as the strand is loaded. Further complications are added to the
Revised 8 October 2015 strand behavior if one or more of the wires break due to strand degradation over time. Although a
Accepted 16 May 2016
detailed investigation on strand behavior is critically important for predicting the capacity of a broken
Available online 11 June 2016
strand as well as developing new monitoring approaches for wire break detection, there is little study
available in the literature on wire breakage in a stressed strand. This paper provides an extensive
Keywords:
investigation on stressing and post-breakage dynamic behavior of a prestressing strand. A finite element
Wire rope
Stranded cable
model is generally useful to study the global strand response, along with many localized phenomena that
Prestressing strand have strong influence on its performance, but are difficult to capture either experimentally or through
Wire breakage closed-form analytical models. Investigations on certain behaviors, such as wire breaks, however, require
Frictional contact a relatively large or even a full-scale model to adequately develop contact and frictional conditions.
Finite element Moreover, such a sizeable model can account for any deviation points and may avoid edge effects.
Cable behavior Consequently, several finite element parameters, such as the load ramp profile and duration, effects of
damping and interwire friction, become critical for an accurate and efficient model. This paper first
presents the use of a parametrized model to study strand behavior and evaluates the effects of these
modeling parameters on strand response; load distribution and redistribution among the wires at the
onset of interwire motion are also considered. The model is then used to simulate wire breakage in a
prestressing strand, so that various aspects of post-breakage response can be examined. Numerical
results show that a linear load ramp or stressing too quickly may lead to an inaccurate axial tension
developed in the strand, whereas the inclusion of nominal mass-based damping has been found effective
in achieving a quasi-static solution at a reasonable computational cost. In addition, the wire break sim-
ulation results indicate that breakage of an outer wire results in greater prestress loss than breakage of
the center wire, which might have important implications for non-destructive wire breakage detection.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction of tension, torsion, flexure, and shear along with multiple nonlinear
phenomena such as interwire motion (the relative movement
Wire strands and ropes often find unique engineering between wires), contact, friction, plasticity, and large deformation.
applications because of their high tensile capacity coupled with Studies on cable mechanics, therefore, have received significant
workable bending flexibility. Accordingly, they are widely used in research attention for well over half a century. Presently, developing
diverse engineering systems ranging from major transportation a better understanding of strand response to wire breaks is critical
structures, such as bridges and aerial cableways, to various hoisting for the development of wire break detection techniques.
equipment like elevators and cranes. The mechanical behavior of Several theoretical models have been proposed in the literature
steel cable, however, is complicated by its intricate geometric pat- to explain the mechanical characteristics of wire strands and ropes,
tern. While being subjected to pretensioning or in-service loading, where strands consist of a layer of wires twisted around a center
a complicated stress-state condition arises that combines the effects wire and ropes consist of strands twisted around a straight core.
Costello and Phillips [1] examined the sensitivity of strand stiffness
to the change in helix angle as loading progresses, as well as inves-
⇑ Corresponding author. Tel.: +1 352 392 9537x1503. tigated its dependency on initial helix angle and end conditions.
E-mail addresses: abm.abdullah@ufl.edu (A.B.M. Abdullah), jrice@ce.ufl.edu (J.A. Their study, however, neglected the effect of friction and wire
Rice), hrh@ce.ufl.edu (H.R. Hamilton), grc@ce.ufl.edu (G.R. Consolazio).

http://dx.doi.org/10.1016/j.engstruct.2016.05.030
0141-0296/Ó 2016 Elsevier Ltd. All rights reserved.
214 A.B.M. Abdullah et al. / Engineering Structures 123 (2016) 213–224

flattening to make the closed-form solution tractable. To obtain the rope with a polymeric fiber core. A FE model of length equal to
static response of complex wire ropes with less computational 1/16 of helical pitch was used to examine the load distribution
effort, Velinsky et al. [2] linearized the nonlinear equations of among wires in the elastic regime as well as the redistribution of
equilibrium. The reduction of effective Young’s modulus with addi- load with the evolution of plastic deformation.
tion of strands was demonstrated by analyzing a rope with an To reduce the computational demand, many of the previously
independent core. Later, Velinsky [3] compared these results with proposed models either make simplifying assumptions or only
nonlinear theory and found them identical in practical load ranges. consider a small segment of rope geometry (partial length and/or
Utting and Jones [4,5] conducted a series of experiments on partial cross section) for analyzing stressed strands. However, a
seven-wire strands under tensile loads and proposed an analytical relatively large model with high mesh resolution is needed to
model that indicates insignificant effects of interwire friction and study various phenomena, such as the strand response after a wire
contact deformation on the overall strand response. Chaplin [6], breakage, so that the contact and frictional conditions may
however, reported that these phenomena, along with other factors, adequately develop. In addition, the representative lengths used
affect the failure mechanisms of wire ropes. Raoof and Kraincanic in existing models are often too short to observe interwire pivoting
[7] considered the effects of interwire friction in their theoretical [24] or stick/slip friction [25,26] and the effects of such phenomena
model and obtained the upper and lower bounds of rope’s effective on load redistribution among wires while stressing. The require-
stiffness that correspond to the no-slip and full-slip condition, ment of a sufficiently large model, combined with high material
respectively. Unlike the classical discrete modeling approach and boundary nonlinearities, make the use of an explicit analysis
[1–3] where each wire is treated as individual helical rod, Jolicoeur scheme a favorable candidate.
and Cardou [8] presented an alternative approach in which each In analyzing complicated contact problems like wire strands
layer of wires is represented by an orthotropic hollow circular and similar structures, explicit time integration has proven
cylinder. This semi-continuous model was applied to several types successful for its computational efficiency, robustness, and solu-
of cable, concluding that this approach tends to produce more tion stability (i.e., lack of convergence difficulty) [17,20,22,27].
satisfactory results for cables with larger number of wires. Elata However, simulating strand stressing in a quasi-static manner
et al. [9] considered the twisted wires in the outer layer of a rope when using an explicit dynamic procedure requires special
and analyzed two extreme kinematic conditions: zero and infinite considerations such as proper selection of loading rate, time varia-
friction between adjacent wires. However, the flexural and tion of applied load, and energy dissipation mechanisms other than
torsional rigidity of wires were neglected. frictional sliding, which become extremely critical in large models.
In general, most of the aforementioned analytical models have The first part of this paper examines the role of these parameters in
made approximations and simplifying assumptions to obtain a achieving an accurate model, and investigates the redistribution of
closed-form solution. Although these models can be used to load among wires at the stick/slip transition.
predict the global response of a cable, they are unable to provide The second part of this paper deals with simulating a wire break
a comprehensive description on many localized phenomena, such in a prestressing strand and investigates post-breakage response.
as yielding along contact lines, uneven bending of outer wires, Although wire breaks in a rope or strand are quite common,
and stress redistribution among wires. With the rapid advancement occurring primarily due to corrosion, to date there have been few
of computing technology, finite element (FE) methods have been attempts to conduct a detailed study on this topic [27–35];
developed over the past few decades to examine these characteris- comprehensive investigation of the behavior of a broken strand is
tics in addition to other critical aspects of wire ropes, such as still an open problem. The use of monostrand tendons in building
microstructural characterization during manufacturing process construction and the recent interest in flexible fillers for
[10] and the mechanisms controlling their ductility [11]. multi-strand post-tensioned tendons in U.S. bridges facilitates
Chiang [12] conducted a FE-based parametric study to show the new monitoring methods for wire break detection [36–38]. Wire
individual and combined effects of different geometric, boundary, breaks in unbonded tendons allow changes in global strand and
and contact conditions on stress response of a strand. By utilizing anchor response, which further highlights the necessity of a
the helical symmetry of geometry and loading, Jiang et al. [13] thorough investigation on strand behavior after the occurrence of
developed a concise FE model of a seven-wire strand. The model any breakage. This paper considers various breakage scenarios,
was further extended [14] to analyze a three-layered 19-wire such as breakages of center or outer wires, varying confinement
strand by updating the constraint equations and boundary conditions, successive wire breaks, and the effect of friction on
conditions. Nawrocki and Labrosse [15] considered different post-breakage response.
interwire motions, namely, sliding, rolling and pivoting, and
showed that pivoting and sliding governs the axial and bending
behavior, respectively. Contrasting the conventional assumption 2. Model geometric and material properties
of contact occurrence only between the center and outer wires,
Jiang et al. [16] demonstrated that the contact also takes place 2.1. Geometric features
between neighboring outer wires.
Erdönmez and Imrak analyzed the behavior of a curved strand Although the construction of wire strands varies in different
[17] and later considered a rope to examine the load distribution parameters, such as wire diameter, helix angle, number of wires,
among wires [18]. Stanova et al. derived parametric equations for group pattern, and lays [39,40], the basic geometry of all these
geometric models of complex wire ropes [19] and implemented strands consists of a straight wire surrounded by a layer of helical
them in a FE program [20]. Zhou and Tian [21] proposed a FE model wires [15]. The cable investigated in this paper is a seven-wire
for single-layered strand based on geometric compatibility and strand (Fig. 1), which is commonly used in prestressed concrete
material elasticity theory. Nodal constraint relations between core (PC) structures. The ASTM A416 Grade 270 strand [41] is made of
and helical wires were obtained for axial tension and bending. six helical wires encasing the center wire. The helix angle and wire
However, the model did not account for the effect of interwire diameters have been chosen such that, in the undeformed configu-
friction or sliding. Kmet et al. [22] investigated a rope deviated ration, each helical wire barely touches its two neighboring helical
over a saddle and observed non-uniform stress distribution among wires in addition to touching the straight center wire [16]. Details of
wires. Fontanari et al. [23] studied the elasto-plastic response of a the geometric and material data are listed in Table 1.
A.B.M. Abdullah et al. / Engineering Structures 123 (2016) 213–224 215

Fig. 1. Seven-wire prestressing strand.

Table 1
Geometric and material properties of a seven-wire prestressing strand [41,42].

Geometry Material
Strand diameter, ds 15.240 mm (0.6 in.) Young’s modulus 198.5 GPa (28,800 ksi)
Center wire diameter, dc 5.150 mm (0.2028 in.) Poisson’s ratio 0.3
Outer wire diameter, do 5.045 mm (0.1986 in.) Yield strain 0.0085
Helical wire pitch, p 190.5 mm (7.5 in.) Yield stress 1687.8 MPa (244.8 ksi)
Helix angle, a 80.46° Rupture strain 0.07
Lay angle, b 9.54° Rupture stress 1857.4 MPa (269.4 ksi)

2.2. Material characterization contact area between wires. Fig. 2 shows that with mesh
refinement, the moduli obtained from FE analyses using different
The material behavior has been characterized by an elastic–plas- mesh densities approach the target modulus. Considering accuracy
tic model with isotropic hardening. Similar assumptions were made (about 98% of target modulus) and computational cost (saving 3/4
in earlier studies and were found reasonable for metallic materials of CPU time by compromising only 0.5% accuracy), the characteris-
[23]. The stress–strain curve for seven-wire prestressing strand in tic radial and circumferential element dimensions were chosen to
the PCI Design Handbook [42] has been considered. The effective be 1/10 of wire diameter, d, and the longitudinal dimension to be
Young’s modulus of the strand, however, is expected to be some- 1/100 of helical pitch, p.
what less than the individual wire modulus because of the change
in helix angle under load [1]. Therefore, the input material modulus 3.2. Boundary conditions and loading
of elasticity was adjusted following Costello’s model [39] so that the
calculated strand modulus matches the target strand modulus in Translation along the strand’s longitudinal axis was restrained
[42]; the difference between wire and strand moduli (207.2 GPa at one end (dead end) but permitted at the other end (live end)
and 198.5 GPa, respectively) was found to be approximately 4.5%. for applying displacement-controlled loading. This loading process
It is noted, however, that because of minor difference between
center and outer wire diameters (Table 1), the size effect on tensile
properties of individual wires [23] is expected to be insignificant
and therefore has not been taken into account.

3. Model development

3.1. Element selection and mesh generation

The geometric volume was discretized with trilinear (8-node)


hexahedral elements that use reduced integration and hourglass
control, which have been successfully implemented in other
studies involving steel cables [12–14,16,17,20,22,23]. A mesh
convergence study was conducted using these elements with the
aim of numerically simulating the effective strand modulus of
elasticity in PCI Design Handbook [42]. For computational effi-
ciency, a relatively short strand length of 190.5 mm (7.5 in.), which
is equal to the length of one helical pitch of the strand (Table 1),
was considered for the mesh convergence study; other modeling
parameters are described in the following section. As anticipated,
a high-resolution mesh was required to effectively model the Fig. 2. Mesh convergence study.
216 A.B.M. Abdullah et al. / Engineering Structures 123 (2016) 213–224

was approximated by imposing a moving boundary along the produce reasonably accurate solutions in other similar applications
strand axis on all wires at the live end. After fully stressing the [16]. Friction was characterized by an isotropic Coulomb friction
strand, the axial translation was prohibited at both terminations model.
to prepare the strand for simulating wire breaks. It is noted that,
in practice, the strands are gripped at the ends by wedges (two- 3.4. Model verification
or three-piece slotted cones to hold the strands in their serrated
teeth) that restrain the elongated strands from returning to their To verify the accuracy of the model, finite element analysis
original lengths. While some slip may occur between the strand (FEA) results were compared with Costello’s theory [39], which is
and wedge at the dead and live end at the beginning of stressing valid only up to the elastic limit. Therefore, to evaluate the analysis
and immediately after removing the stressing jack, respectively, results beyond material yielding, the complete stress–strain curve
the strands are ultimately locked and held in place by the wedges. obtained from FEA was also compared with the curve provided in
In addition, although the wedge grips only the outer wires, the PCI Design Handbook [42]. The minor mismatch between the
applied load is transmitted to the center wire through friction analytical and FEA curve (difference in moduli less than 2.5%), as
under high normal pressure exerted by the wedging mechanism. can be seen in Fig. 4, is in part because of mesh resolution
Thus, the gripping of the strands by the wedges was modeled by (Fig. 2) and also due to the fact that contact deformation was
directly restraining all the wires from moving along the strand’s neglected in Costello’s model.
axis. To prevent unwinding of helical wires, the ends must be con-
strained from axial rotation; this was achieved by restraining the 4. Sensitivity analysis of FE parameters
circumferential motion (U(t) = Fixed). Furthermore, radial contrac-
tion (during stressing) and expansion (after breakage) of wires due This section investigates the sensitivity of strand response to
to Poisson effect were allowed at both ends (U(r) = Free). The several FE parameters, namely, load ramp profile and duration,
details of imposed boundary and loading conditions are shown in friction model, and damping. Appropriate parameters have been
Fig. 3. determined for an accurate and efficient model to be used in
analyzing strand behavior as well as simulating wire breakages
3.3. Interwire friction in the subsequent sections. The modeling parameters used in
different studies are shown in Table 2.
The FE model must also address friction at the contact points
between wires. Therefore, the model is required to account for 4.1. Load ramp profile
all the complex mechanical phenomena originating from contact
nonlinearities. The wire-to-wire contact interactions were The time variation of prescribed displacement at the live end,
considered as deformable-to-deformable contact pairs. Finite slid- which is the strand’s total elongation, has a notable effect on strand
ing formulation was used to account for possible large interwire response during simulation of prestressing. A linear displacement
motion after wire breaks as well as at the end of sticking phase ramp with time (Fig. 5a) produces an oscillatory response (Fig. 5b)
while stressing. Surface-to-surface contact discretization was used because of sudden application of velocity, which also causes the
as it reduces the likelihood of large localized penetrations of kinetic energy to fluctuate throughout the ramp duration (Fig. 6).
master nodes into slave surface and also reduces the sensitivity The time lag of response (tL ) observed in the inset of Fig. 5b (time
of results to master/slave roles [43]. Penalty functions were used vs. strand prestressing force plot) is the time required by the stress
to enforce the impenetrability requirement. Although the exact ful- wave to traverse the strand length. In estimating the transit time of
fillment of the constraint condition is compromised, this approach propagated wave, the dilatational wave speed was determined by
does not increase the number of unknowns (hence computational computing simplified effective moduli from the material’s consti-
demand) like other algorithms, such as Lagrange multiplier or tutive response [43]. A quasi-static response is obtained (indicated
augmented Lagrangian formulations, but has been shown to by the elongation vs. strand prestressing force plot in Fig. 5b),

Fig. 3. Boundary conditions and loading.


A.B.M. Abdullah et al. / Engineering Structures 123 (2016) 213–224 217

Fig. 4. Stress–strain curves for strand. Fig. 6. Kinetic energy history for linear and smooth load ramp.

however, if the displacement is ramped up such that velocity


(dU=dt) is zero at the beginning, then increases smoothly until over the loading process with a peak at the middle, confirming the
the middle of the step, followed by decreasing over the rest of consistency of the applied procedure. It is noted that the ramp
the time and finally return to zero (Fig. 5a) [43]. Following the speed (or loading rate) was kept sufficiently slow during both lin-
velocity profile, the kinetic energy history (Fig. 6) varies smoothly ear and smooth load ramp to avoid any unwanted dynamic effect.

Table 2
Parameters for analyzing strand response during stressing.

Model ID Section Variable Length (mm) l Load ramp profile n (%) Load ramp duration
S1 3.1 Mesh resolution 190.5 0.1 Smooth 10 10Tn
S2 4.1 Load ramp profile 190.5 0.1 – 0 10Tn
S3 4.2 Load ramp duration 190.5 0.1 Smooth 10 –
S4 4.3 Friction model 952.5 – Smooth 10 10Tn
S5 4.4 Damping 952.5 0.1 Smooth – 10Tn
S6 5 Localized response 190.5 0.1 Smooth 10 10Tn
S6 6 Load redistribution among wires 190.5 0.1 Smooth 10 10Tn
S7 6 Load redistribution among wires 952.5 0.1 Smooth 10 10Tn

l: Friction coefficient; n: Fraction of mass critical damping (damping ratio) corresponding to the fundamental longitudinal mode of vibration; Tn: natural period of the
fundamental longitudinal mode of vibration (estimated Tn for 190.5 mm and 952.5 mm long strands are 0.15 ms and 0.75 ms, respectively [47]).

Fig. 5. Time variation of applied loads: (a) ramp profile; (b) developed force vs. elongation.
218 A.B.M. Abdullah et al. / Engineering Structures 123 (2016) 213–224

Fig. 7. Strand prestressing force vs. load ramp duration.

Fig. 8. Exponential decay friction model.

4.2. Load ramp duration

In practice, the actual time taken to complete the stressing


operation of a prestressing strand is relatively long (at least in the
order of few seconds). Stressing a long strand in such a physical
timescale, however, is often not feasible in FEA due to very small
time increment (generally, in the order of nanoseconds in this type
of analysis) caused by the extra fine mesh. Therefore, a parametric
study has been conducted with various ramp durations to obtain
an economical solution while not being significantly affected by
inertial effects. Analysis results show that a load ramp duration of
at least three times the natural period of the fundamental longitudi-
nal (axial) mode of vibration, Tn, is adequate for this particular anal-
ysis to achieve a static response (Fig. 7). A higher loading rate results
in non-uniform stress distribution along the strand, which indicates
the inadequacy of the analysis. Moreover, the higher-than-static
forces suggest the presence of dynamic effects.

4.3. Friction model


Fig. 9. Occurrence of interwire slip.
In the literature, apart from theoretical models that consider
either or both the two extreme cases of zero and infinite friction
[1,3,7–9], the interwire frictional effect has been predominantly As expected, Fig. 10 shows that strand responses from the models
modeled with a constant friction coefficient (considered for other with a constant friction coefficient (l ¼ 0:1) and an exponential
analyses reported in this paper) lying between 0.1 and 0.2 decay friction model (ls ¼ 0:2; lk ¼ 0:1) mostly overlap, indicating
[13,14,16–18,20,22,23]. In this section, these two values are taken that the two responses are quite identical, including the onset of
as lower and upper bounds of friction, respectively, and are defined slip. Thus, Coulomb friction model with a constant friction coeffi-
as kinetic (lk ) and static (ls ) friction coefficients with an cient has been found adequate for the range of friction coefficients
exponential decay model (Fig. 8) [43]. [13,14,16–18,20,22,23] generally considered in analyzing strand
Fig. 9 illustrates the transition from interwire sticking state to response. A constant friction equal to the kinetic friction coefficient
slipping as the loading progresses. The onset of interwire slip can is, therefore, used in the rest of the analyses for simplicity.
be also confirmed by the sliding energy history. It is noteworthy
that this stick/slip transition was not obvious in shorter strand
models (S1, S2, or S3 in Table 2), which appears to be because of 4.4. Damping
edge effects, but became evident in relatively long strand of length
equal to five times the helical pitch (952.5 mm). As observed in previous plots (Figs. 5b, 6, and 9), linear load
The final prestressing force at the end of stressing, as well as the ramp or stick/slip transition during stressing induces dynamic
post-breakage response, however, are mostly governed by effects in the model, which resulted in an oscillatory response.
the slipping portion of frictional condition. Although shifting of Although the vibration energy was dissipated through frictional
the onset of slip is possible, the magnitude of friction coefficient sliding, it required significant computation time for the oscillation
in the sticking phase is not expected to significantly affect the to die out. This is because the sliding energy was relatively small
strand response because of the lack of interwire motion [39,40]. [24]. Therefore, an additional energy dissipation mechanism was
A.B.M. Abdullah et al. / Engineering Structures 123 (2016) 213–224 219

Fig. 10. Constant friction coefficient vs. exponential decay model.


Fig. 12. Addition of mass damping to damp the oscillation originating from
interwire slip.

introduced utilizing mass proportional damping to achieve a static


response in a tractable timescale. The analysis results show that an developed at the end of stressing. Similar trend of strain
addition of nominal damping (10% of the critical mass damping distribution was obtained in a study on Warrington-Seale rope
corresponding to the fundamental longitudinal mode of vibration) [23]. The Mises stress distribution in the wires is illustrated in
was effective in damping the oscillation (Figs. 11 and 12). Fig. 13b. As expected, the maximum stress was observed along
the contact lines between the center and outer wires. The model
also accounted for other critical phenomena, such as energy dissi-
5. Evolution of strand’s localized response with stressing
pation through frictional sliding, which has been discussed in the
following section.
Apart from global strand response (i.e., prestress loss due to
wire breaks, described later), the model also provides insights into
several localized phenomena such as contact pressure, frictional 6. Load redistribution among wires
sliding, and yielding. The contact pressure developed along six
helical contact lines on the straight center wire and along a single The center wire has greater axial stiffness than the outer wires
contact line on helical outer wires. The contact pressure increased because of its larger diameter (Table 1) as well as the frictional
with strand elongation, which introduced plasticity in the wires contact provided by the surrounding wires. In addition, although
along contact lines when the strand was elongated about 0.75% the displacement of the outer wires along their own axes is larger
of its original length. The material yielding started developing than that in the center wire, which is reduced proportionally to the
around the mid-length region and then gradually extended toward updated lay angle (Fig. 1), the axial strain in outer wires is smaller
the terminations. Fig. 13a shows the equivalent plastic strain than that in the center wire [1]. This is because, when the strand is

Fig. 11. Effect of damping on linear load ramp: (a) developed force; (b) kinetic energy.
220 A.B.M. Abdullah et al. / Engineering Structures 123 (2016) 213–224

Fig. 13. Strand response at 0.8% elongation: (a) equivalent plastic strain; (b) Mises stress (ksi).

under uniaxial load, the center wire displaces along the strand axis
while the outer wires progressively align with the axis [23]. As a
result, the center wire carries a larger share of axial force than
the outer wires, provided that there is no significant interwire
motion. Such a condition is observed in a relatively short strand
model, i.e., model S6 in Fig. 14, where the relative motion between
the wires is restricted by edge effects.
The interwire motion, however, becomes more pronounced in a
relatively long strand (model S7) when the wires experience a post-
slip state of friction. As expected, the center wire initially carries a
greater share of axial force as long as the material remains elastic
and there is no considerable relative motion (Fig. 15). After the
stick/slip transition occurs, a portion of the system energy is dissi-
pated through frictional sliding. Because the center wire has more
contact areas than the outer wires, the sliding energy dissipation
associated with the center wire is higher than the outer wires. In
addition, the onset of yielding along the contact helices primarily
affects the center wire, which contributes further to redistribute
the load to the outer wires [23]. Thus, the initial load distribution
Fig. 14. Comparison of load distribution among wires in a long- and short-strand. is finally reversed and the outer wires carry a larger share (Table 3).

Fig. 15. Load redistribution among wires with the onset of slip.
A.B.M. Abdullah et al. / Engineering Structures 123 (2016) 213–224 221

Table 3
Final load distribution among wires.

Model ID (Table 2) Total load (kN) Center wire Outer wire


S6 209.5 31.3 kN 14.94% 29.7 kN 14.18%
S7 209.0 29.0 kN 13.90% 30.0 kN 14.35%

7. Wire break simulation and post-breakage response

This section covers simulations of a center wire break, an outer


wire break, a wire break in a confined strand, and successive
multiple wire breaks (Table 4). Such investigation has practical
significance because it enables the analysis of the impact of wire
breaks on structural performance and can inform the development
of wire breakage detection methods. Of particular interest is the
change in prestressing force that occurs when wires break.

7.1. Wire break simulation procedure


Fig. 16. Typical stressing profile and post-breakage response.

The strand was first stressed to 80% of its ultimate strength


(0.80 Fu, where Fu represents the specified tensile strength of pre-
stressing steel) [41,44,45]. Because a wire break commonly occurs Table 5
Prestress loss due to wire breaks.
in an aged strand due to corrosion, the load was then ramped down
to 0.60 Fu to obtain the final effective force, excluding the prestress Model ID (Table 4) Prestress loss (%)
losses, such as elastic shortening, shrinkage, and creep of concrete, B1 6.6
relaxation of steel, anchor set [46] that commonly occur in PC B2 17.4
structures. The complete stressing history was necessary to B3 15.3
account for the plasticity that forms in the contact areas between
the wires when the strand experiences higher stresses. Once the
pretensioning step was completed, the stiffness and strength of a 7.2. Post-breakage response
target element group, consists of the elements in a wire located
at a position where the break was intended, were artificially 7.2.1. Center and outer wire breaks
reduced to locally increase the effective strain within these An analysis was conducted to compare the load loss due to
elements. Thus, the length over which damage was imposed was breakages in center and outer wires. The results show that break-
equal to the dimension of an element along the wire axis. To intro- age in an outer wire caused approximately 17.4% prestress loss in
duce damage in the element group, a certain value of equivalent the strand, whereas breakage in the core wire resulted in only 6.6%
plastic strain, around 5.15% in this simulation, was set as damage loss (Table 5). This greater load loss due to an outer wire break was
initiation criterion. Once this criterion was reached, the material expected because the center and adjacent wires partially lost their
was progressively led to failure following a linear law for damage frictional contact area and radial pressure in addition to lose of the
evolution with effective plastic displacement [43]. The failure broken wire’s share of the total force (Table 3). In addition, the
mechanism was then completed when the effective plastic smaller load loss in case of the core wire break is attributed to
displacement reached a predetermined ultimate value (6.15%) the fact that it was under much higher radial pressure than the
and finally, the fully degraded elements were deleted from the outer wires, which resulted in greater frictional resistance.
discretized geometry. It can be noted that the low-relaxation pre-
stressing strand is typically stretched and annealed to reduce
residual stresses induced by the cold-drawing process. Moreover, 7.2.2. Outer wire breaks in a confined strand
the generated wire break was not caused by strand rupture due In practice, a mono-strand tendon is generally placed in a plastic
to overstressing. Thus the effect of residual stresses left by the sheath for corrosion protection. Similarly, in a multi-strand system,
manufacturing process is expected to be marginal on the overall a bundle of strands is usually contained inside a duct. The plastic
strand response, and therefore has not been taken into account. sheath in mono-strand tendons and the presence of duct and other
Fig. 16 shows a complete stressing profile with a typical strands in a multi-strand setting may potentially limit a wire’s
post-breakage response. lateral movement after breakage. This situation has been simulated

Table 4
Wire breakage simulation matrix.

Model ID Section Broken wire Length (mm) Break location from dead end (mm) Confinement condition Friction coefficient, l
B1 7.2.1 Center 952.5 857.25 Bare 0.1
B2 7.2.1 Outer 952.5 857.25 Bare 0.1
B3 7.2.2 Outer 952.5 857.25 Confined 0.1
B4 7.2.3 Center 952.5 857.25 Bare -Varies-
B5 7.2.3 Outer 3352.8 3048.0 Bare -Varies-
B6 7.2.4 Outer 19,812.0 -Varies- Bare 0.3
B7 7.2.5 Outer (multiple) 952.5 857.25 Bare 0.1
222 A.B.M. Abdullah et al. / Engineering Structures 123 (2016) 213–224

between the cylinder and the strand is considered frictionless and


no confining pressure has been applied to the cylinder. As listed
in Table 5, a wire break in confined condition contains greater
prestressing force after breakage compared to the bare condition.
This is because the broken wire is restricted from being separated
from its neighboring wires in confined condition, which results in
greater frictional resistance to prestress loss.

7.2.3. Center and outer wire breaks in a strand with various friction
coefficients
Force loss due to wire breakage in an aged, partially corroded
strand is expected to be different from that in a new, uncorroded
strand. A parametric study was conducted on strands with various
friction coefficients (using Coulomb friction model with a constant
friction coefficient), in which a lower coefficient represented a
newer strand, whereas a higher coefficient characterized a corroded
strand. The study also helped examine the effect of lubrication on
Fig. 17. FE model of strand with radial confinement (rigid elements). strand response in the presence of different filler materials used
in post-tensioning ducts. Fig. 18 shows the remaining prestressing
force after an outer wire breakage in an approximately 3.4 m-long
strand (Table 4) with different friction coefficients. A similar study
was performed with shorter strand length (952.5 mm) for core wire
breaks. Force loss was significantly less due to breakage in core wire
even with shorter length, which is because the core wire was under
much higher radial pressure exerted by the helical wires. The
results indicate that the total prestressing force may remain almost
unaffected by a breakage in the core wire in an aging strand. Fig. 18
illustrates mild nonlinearity in the relationship between force loss
and friction coefficient, for both center and outer wire breaks. This
nonlinearity appears to be due to the effect of friction on formation
of birdcaging (Fig. 19), which was formed at about 2.4 m away from
the wire break in model B5 with a friction coefficient of 0.1, as well
as the loss of contact between broken wire and the rest of the strand
around break location.

7.2.4. Outer wire breaks at different locations in a long strand


A parametric study was conducted where wire breaks were
simulated at different locations along the length of a nearly
20 m-long strand. A coarser mesh (d/4  d/4  p/37.5 or
Fig. 18. Effect of friction coefficient on remaining strand prestressing force. 1.27 mm  1.27 mm  5.08 mm) was used to keep the problem
size tractable, and a relatively high friction coefficient (l ¼ 0:3)
was used to emphasize the effect of break location on prestress loss
by modeling a cylindrical container around the strand with the in a partially corroded strand.
cylinder barely touching the strand in an unloaded state (Fig. 17). As expected, Fig. 20 shows that the prestressing force was less
To minimize additional computational expense, the cylindrical affected when the breakage occurred away from the dead end. It
geometry has been discretized with rigid elements. The interaction is noteworthy that parameters such as birdcage location and size,

Fig. 19. Birdcage formation and broken wire separation.


A.B.M. Abdullah et al. / Engineering Structures 123 (2016) 213–224 223

Table 6
Frequency decrease with wire breaks.

Breakage condition Frequency


Hz Decrease (%)
Unbroken strand 754.1 –
First wire break 719.6 4.6
Third wire break 612.5 18.8

along with the length around the breakage location over which the
broken wire remains completely separated from the strand, vary
with the break length because the strain energy stored by wires
changes along the length. As a result, the force loss was non-
proportional (somewhat nonlinearly related) to the break length
due to complex interaction of multiple factors, such as, friction,
birdcaging as well as the separation length around break location.

7.2.5. Successive wire breaks


After analyzing various wire break conditions and their effects
on prestressing force in the earlier sections, this section investigates
Fig. 20. Remaining prestressing force due to wire breaks at different locations in a the dynamic post-breakage behavior of strand. Such investigation
long strand. can be useful in developing new breakage detection methods that
continuously monitor the strand response at the end anchors to
capture any change in strand dynamics. A multiple wire break con-
dition (Fig. 21) has been considered and natural frequencies of the
contributing vibration modes were extracted from time history
after each wire breaks. Fig. 22 shows the magnitude spectra of
the vibration response of the strand after the first and third wire
breaks. The obtained frequencies (determined by peaks in the spec-
tra) were compared (Table 6) with the frequency of unbroken
strand estimated through an Eigen analysis as well as an analytical
prediction [47]. It is observed that the frequencies, as well as the
respective magnitudes, decrease with successive wire breaks. This
decrease of frequency is because of the reduced tension in the
strand and also because of the decreasing strand stiffness with wire
breaks due to cumulative loss of broken wire cross-sectional area
along with reduction in confinement. In addition, the damping
(estimated by half-power bandwidth) tends to increase with the
number of wire breaks because of greater interactions among the
Fig. 21. Multiple wire breaks. broken wires.

8. Conclusions

A FE model of prestressing strand has been developed for a


detailed analysis of its mechanical behavior. The analysis facilitates
a comprehensive understanding of both static and dynamic
response of the strand, and captures useful information on strand
behavior for developing wire break detection methods. The model
considers an accurate representation of material, geometric, and
frictional conditions that enables the investigation on interwire
motion during stressing as well as the strand response after a wire
breakage. A sensitivity study of several FE parameters, such as load
ramp profile, load ramp duration, friction model (Coulomb friction
model with a constant friction coefficient and an exponential decay
friction model with both static and kinetic friction coefficients) and
damping, has been conducted first to determine the appropriate
parameters for an efficient model. After validating the analysis
results with available analytical works in the literature, the model
was then used to conduct a detailed investigation on load distribu-
tion among wires, followed by various parametric studies on wire
breaks.
The analysis results show that a linear load ramp profile or
ramp duration less than three times the fundamental longitudinal
period of the strand induces unwanted dynamic effects into the
Fig. 22. Frequency shift with successive wire breaks. system. The results also demonstrate that the interwire friction
224 A.B.M. Abdullah et al. / Engineering Structures 123 (2016) 213–224

can be adequately modeled with a constant friction coefficient, [10] Fontanari V, Bulf U, Benedetti M. Numerical analysis of the rolling process of
shaped wires for locked steel ropes. J Mater Process Technol 2005;170
which is equal to the kinematic friction coefficient. The presence
(1):97–107.
of inertial effects was also observed due to stick/slip transition [11] Phelippeau A, Pommier S, Tsakalakos T, Clavel M, Prioul C. Cold drawn steel
and the addition of nominal mass damping has been found wires—processing, residual stresses and ductility—Part I: Metallography and
effective in obtaining a static response. The model illustrates the finite element analyses. Fatigue Fract Eng Mater Struct 2006;29(3):201–7.
[12] Chiang YJ. Characterizing simple-stranded wire cables under axial loading.
influence of interwire motion on strand response, especially on Finite Elem Anal Des 1996;24(2):49–66.
the load distribution among wires. Though the center wire carries [13] Jiang WG, Henshall JL, Walton JM. A concise finite element model for three-
a larger share of total load than the outer wires before slippage, layered straight wire rope strand. Int J Mech Sci 2000;42(1):63–86.
[14] Jiang WG, Yao MS, Walton JM. A concise finite element model for simple
this distribution has been found to be reversed after the occurrence straight wire rope strand. Int J Mech Sci 1999;41(2):143–61.
of stick/slip transition, when the outer wires begin to carry a larger [15] Nawrocki A, Labrosse M. A finite element model for simple straight wire rope
share. Furthermore, the post-breakage response demonstrates that strands. Comput Struct 2000;77(4):345–59.
[16] Jiang WG, Warby MK, Henshall JL. Statically indeterminate contacts in axially
an outer wire break causes significantly higher force loss (17.4% loaded wire strand. Eur J Mech-A/Solids 2008;27(1):69–78.
loss) compared to the center wire (6.6% loss), while a confined [17] Erdönmez C, Imrak CE. Modeling and numerical analysis of the wire strand. J
strand loses less force due to a wire break than a bare strand. Mild Naval Sci Eng 2009;5(1):30–8.
[18] Imrak CE, Erdönmez C, Gumussuyu I, Maslak I. On the problem of wire rope
nonlinearities have been observed in strand response with friction model generation with axial loading. Math Comput Appl 2010;15(2):259–68.
and break location. The analysis results also indicate correlation [19] Stanova E, Fedorko G, Fabian M, Kmet S. Computer modelling of wire strands
between successive wire breaks and change in dynamic character- and ropes Part I: Theory and computer implementation. Adv Eng Softw
2011;42(6):305–15.
istics; the strand’s fundamental longitudinal natural frequency
[20] Stanova E, Fedorko G, Fabian M, Kmet S. Computer modelling of wire strands
decreased by 18.8% after three successive wire breaks. and ropes part II: Finite element-based applications. Adv Eng Softw 2011;42
The research presented in this paper primarily considers a (6):322–31.
straight strand with helical wires under an external load applied [21] Zhou W, Tian HQ. A novel finite element model for single-layered wire strand. J
Cent South Univ 2013;20:1767–71.
along the strand axis. Further investigation is to be made on [22] Kmet S, Stanova E, Fedorko G, Fabian M, Brodniansky J. Experimental
simulating wire breaks in a deviated strand under a realistic radial investigation and finite element analysis of a four-layered spiral strand bent
pressure in addition to the axial tension. Wire breakage in a over a curved support. Eng Struct 2013;57:475–83.
[23] Fontanari V, Benedetti M, Monelli BD. Elasto-plastic behavior of a Warrington-
multi-strand tendon would be interesting as well to obtain a more Seale rope: experimental analysis and finite element modeling. Eng Struct
practical representation of interaction of the broken wire with 2015;82:113–20.
surrounding strands and its effects on overall response. The [24] Labrosse M, Nawrocki A, Conway T. Frictional dissipation in axially loaded
simple straight strands. J Eng Mech 2000;126(6):641–6.
observation on frequency shifts with successive wire breaks [25] Huang X, Vinogradov OG. Dry friction losses in axially loaded cables. Struct
appears to be promising for strand breakage detection; however, Eng Mech 1996;4(3):330–44.
more investigation is necessary to determine its feasibility as a [26] Huang XL, Vinogradov OG. Extension of a cable in the presence of dry friction.
Struct Eng Mech 1996;4(3):313–29.
potential wire break detection criterion. An extensive experimental [27] Waisman H, Montoya A, Betti R, Noyan IC. Load transfer and recovery length in
program has been designed to investigate the recovery length or parallel wires of suspension bridge cables. J Eng Mech 2010;137(4):227–37.
prestress loss due to wire breaks. Multiple wire cuts are intended [28] Chien CH, Costello GA. Effective length of a fractured wire in wire rope. J Eng
Mech 1985;111(7):952–61.
in both bare and confined strands as well as in internal and external
[29] Gjelsvik A. Development length for single wire in suspension bridge cable. J
tendons with the presence of deviators and flexible fillers. The Struct Eng 1991;117(4):1189–200.
experimental findings are to be subsequently used to validate and [30] Raoof M. Wire recovery length in a helical strand under axial-fatigue loading.
extend the numerical model reported in this paper. Int J Fatigue 1991;13(2):127–32.
[31] Raoof M, Huang YP. Wire recovery length in suspension bridge cable. J Struct
Eng 1992;118(12):3255–67.
[32] Raoof M, Kraincanic I. Determination of wire recovery length in steel cables
Acknowledgements and its practical applications. Comput Struct 1998;68(5):445–59.
[33] MacDougall C, Bartlett FM. Mechanical model for unbonded seven-wire
tendon with symmetric wire breaks. J Eng Mech 2005;131(12):1239–47.
The presented work is a part of the research project Replaceable [34] MacDougall C, Bartlett FM. Mechanical model for unbonded seven-wire
Unbonded Tendons for Post-Tensioned Bridges, funded by the Florida tendon with single broken wire. J Eng Mech 2006;132(12):1345–53.
Department of Transportation. The authors also acknowledge the [35] MacDougall C, Bartlett FM. Tests of unbonded seven-wire tendon with broken
outer wires. ACI Struct J 2003;100(5).
support from graduate research fellowship at the University of [36] Abdullah ABM, Rice JA, Hamilton HR. Wire breakage detection using relative
Florida. strain variation in unbonded posttensioning anchors. J Bridge Eng 2014;20(1).
[37] Abdullah ABM, Rice JA, Hamilton HR. A strain-based wire breakage
identification algorithm for unbonded PT tendons. Smart Struct Syst 2015;16
(3).
References [38] Abdullah ABM, Rice JA, Hamilton HR. A damage detection model for unbonded
post-tensioning tendons based on relative strain variation in multi-strand
[1] Costello GA, Phillips JW. Effective modulus of twisted wire cables. J Eng Mech anchors. In: Proc. SPIE smart structures/NDE 2014, vol. 9061. San Diego, CA;
Div 1976;102(1):171–81. 2014.
[2] Velinsky SA, Anderson GL, Costello GA. Wire rope with complex cross sections. [39] Costello GA, Costello GA. Theory of wire rope. New York: Springer-Verlag;
J Eng Mech 1984;110(3):380–91. 1990.
[3] Velinsky SA. General nonlinear theory for complex wire rope. Int J Mech Sci [40] Feyrer K. Wire ropes. Berlin: Springer-Verlag Berlin Heidelberg; 2007.
1985;27(7):497–507. [41] ASTM. Standard specification for steel strand, uncoated sevenwire for
[4] Utting WS, Jones N. The response of wire rope strands to axial tensile loads— prestressed concrete. A416, West Conshohocken, PA; 2006.
Part I. Experimental results and theoretical predictions. Int J Mech Sci 1987;29 [42] PCI Industry Handbook Committee. PCI design handbook: precast and
(9):605–19. prestressed concrete; 2010.
[5] Utting WS, Jones N. The response of wire rope strands to axial tensile loads— [43] Dassault Systemes Simulia Corporation. Abaqus 6.12 documentation; 2012.
Part II. Comparison of experimental results and theoretical predictions. Int J [44] ACI. Building code requirements for structural concrete (ACI 318-14) and
Mech Sci 1987;29(9):621–36. commentary (ACI 318R-14). American Concrete Institute; 2014.
[6] Chaplin CR. Failure mechanisms in wire ropes. Eng Fail Anal 1995;2(1):45–57. [45] AASHTO. AASHTO LRFD bridge design specifications. 5th ed. Washington, DC;
[7] Raoof M, Kraincanic I. Analysis of large diameter steel ropes. J Eng Mech 2010.
1995;121(6):667–75. [46] Kelley GS. Prestress losses in post-tensioned structures. PTI technical note
[8] Jolicoeur C, Cardou A. Semicontinuous mathematical model for bending of #10; Sept. 2000.
multilayered wire strands. J Eng Mech 1996;122(7):643–50. [47] Meirovitch L. Analytical methods in vibrations. London: Macmillan Company;
[9] Elata D, Eshkenazy R, Weiss MP. The mechanical behavior of a wire rope with 1967.
an independent wire rope core. Int J Solids Struct 2004;41(5):1157–72.